Thanks so much! I'm not sure how I didn't catch this sooner. Change LGTM. I'll send an updated integration test for the other diff.

I think GPU function support for EmitC may be better off as a GPUToEmitC conversion pass where gpu.func gets converted to func.func. This would subsume the pass you added as well as the changes to the emitter.

In D140644#4052291, @NathanielMcVicar wrote:

This appears to have broken the Windows buildbot. https://lab.llvm.org/buildbot/#/builders/13/builds/30769/steps/6/logs/stdio

C:\buildbot\mlir-x64-windows-ninja\llvm-project\mlir\lib\Conversion\GPUToNVVM\LowerGpuOpsToNVVMOps.cpp(209) : error C2220: the following warning is treated as an error
C:\buildbot\mlir-x64-windows-ninja\llvm-project\mlir\lib\Conversion\GPUToNVVM\LowerGpuOpsToNVVMOps.cpp(209) : warning C4715: '<lambda_7426c7b91625e1020521dab1f464b370>::operator()': not all control paths return a value

C:\buildbot\mlir-x64-windows-ninja\llvm-project\mlir\lib\Conversion\GPUToROCDL\LowerGpuOpsToROCDLOps.cpp(149) : error C2220: the following warning is treated as an error
C:\buildbot\mlir-x64-windows-ninja\llvm-project\mlir\lib\Conversion\GPUToROCDL\LowerGpuOpsToROCDLOps.cpp(149) : warning C4715: '<lambda_fea91019496d34fd405e9fcb5563e11d>::operator()': not all control paths return a value

Fix commit message.

Fix trailing whitespace.

Add some tests for the standalone "gpu-lower-memory-space-attributes" pass.

LGTM!

F32 needs to be extended to f64. If you add a test you will see it fails.

Address all comments

The code for the pass that is introduced here is being exercised in the gpu-to-llvm conversions, but I still need to add tests for the pass in isolation.

Can you add some more detail to the commit message? Per our offline discussion, the issue became apparent when the .x4 variant is used with transpose = true. The changes in the tests are mostly NFC with the exception of m16n16k16_mmasync16816_fp16_f16_row_row_row, where the rowB/colB affine maps were incorrect. For the other .x2 test cases, the new affine maps introduced here and the old affine maps in the CHECK statements are equivalent because only the first 16 thread ids in the warp matter.

Fix missing change of nvgpu metadata type to vector<2xi16>

Looks good to me.

Fix typo

Address comments

Address comment, use strategy of rank-reducing extract slice op on the source of the collapse shape.

I implemented the suggestion today, and it works well, but will put up the cleaned up version early next week.

Looks great, made a few minor comments. Thanks!

Fix missing newline

Thanks!

Closing in favor of introducing interface. Here is the draft diff: https://reviews.llvm.org/D134393

Seems like the utility function mergeOffsetsSizesAndStrides should be added to ViewLikeInterface or Dialect/Utils because it has greater scope. The primary use that comes to mind is that the memref "fold alias ops" (used to be "fold subviews") essential does the same thing (SubView and ExtractSlice are both implement the OffsetSizeAndStrideOpInterface). Therefore, it seems like an opportunity for code de-duplication.

Remove unused changes to RegionUtils.h

Address remaining comments.

Address more comments.

Address reviewer comments

Fix missing doc

Cleanup fix some mistakes and inefficiencies.

More general solution

Handle uses of nested values defined above in same stage / different stage.
May need some additional cleanup.

Fix formatting issues.

Rebase on D133523, resolve linalg dependency issue.

This is needed to address circular dependency blocking https://reviews.llvm.org/D129699.

Depends on https://reviews.llvm.org/D133523 to resolve circular dependency with ViewLikeInterface

In D129699#3768090, @mehdi_amini wrote:

There is an issue of circular dependency here: you're introducing a dependency from Dialect/Utils/ to the ViewLikeInterface, but it already depends on Dialect/Utils

Windows bot seems to be failing for unrelated reason

Rebase

Thanks for adding this.

Fix typo in doc

Fix typos.

Expand the documentation for the main class that is added. I checked the result of the Doxygen build to make sure it looks correct with relevant hyperlinks and headings correctly generated.

Rebase

Address comments